Samsung Launches 3rd-generation (16GB) HBM2E
HBM2E stacks eight 16Gb DRAM dies to achieve 16GB package capacity and ensures a stable data transfer speed at 3.2Gbps
Samsung Electronics, the world leader in advanced memory technology, today announced the market launch of ‘Flashbolt’, its third-generation High Bandwidth Memory 2E (HBM2E). The new 16-gigabyte (GB) HBM2E is uniquely suited to maximize high performance computing (HPC) systems and help system manufacturers to advance their supercomputers, AI-driven data analytics and state-of-the-art graphics systems in a timely manner.
“With the introduction of the highest performing DRAM available today, we are taking a critical step to enhance our role as the leading innovator in the fast-growing premium memory market,” said Cheol Choi, executive vice president of Memory Sales & Marketing at Samsung Electronics. “Samsung will continue to deliver on its commitment to bring truly differentiated solutions as we reinforce our edge in the global memory marketplace.”
Ready to deliver twice the capacity of the previous-generation 8GB HBM2 ‘Aquabolt’, the new Flashbolt also sharply increases performance and power efficiency to significantly improve next-generation computing systems. The 16GB capacity is achieved by vertically stacking eight layers of 10nm-class (1y) 16-gigabit (Gb) DRAM dies on top of a buffer chip. This HBM2E package is then interconnected in a precise arrangement of more than 40,000 ‘through silicon via’ (TSV) microbumps, with each 16Gb die containing over 5,600 of these microscopic holes.
Samsung’s Flashbolt provides a highly reliable data transfer speed of 3.2 gigabits per second (Gbps) by leveraging a proprietary optimized circuit design for signal transmission, while offering a memory bandwidth of 410GB/s per stack. Samsung’s HBM2E can also attain a transfer speed of 4.2Gbps, the maximum tested data rate to date, enabling up to a 538GB/s bandwidth per stack in certain future applications. This would represent a 1.75x enhancement over Aquabolt’s 307GB/s.
Samsung expects to begin volume production during the first half of this year. The company will continue providing its second-generation Aquabolt lineup while expanding its third-generation Flashbolt offering, and will further strengthen collaborations with ecosystem partners in next-generation systems as it accelerates the transition to HBM solutions throughout the premium memory market.
Leaked photos confirm rumors about Samsung Galaxy S20 - 01/13/2020 10:24 AM
Leaked photos seem to confirm that Samsung will not reveal the Galaxy S11 series on 11 February, but will switch to the Galaxy S20 series. It concerns three models: Samsung Galaxy S20, S20 Plus and S2...
Blackout at the Samsung factory should not affect DRAM prices in the first quarter of 2020 - 01/10/2020 11:12 AM
Earlier last week we reported that Samsung had a power outage at a a factory for semiconductors, the outage that lasted one minute put the plant down for over a day and brought production of DRAM and...
Samsung 980 PRO M.2 gets PCIe Gen4 and performance up to 6,500 MB/s - 01/09/2020 12:52 PM
Samsung will be releasing its first PCIe 4.0 SSDs dubbed the 980 Pro, and on that PCIe Gen 4.0 link it achieves read speeds of up to 6.5 GB/s, there will be three models. ...
Samsung T7 Touch Portable SSD (With a Fingerprint Sensor) is Out - 01/09/2020 10:46 AM
Samsung announced the release of its latest external storage device – the Portable Solid State Drive (SSD) T7 Touch – which combines sleek, compact design with the fastest transfer speeds currentl...
Samsung Electronics Unveils 2020 QLED 8K TV at CES - 01/06/2020 10:27 AM
Samsung Electronics introduced today its next-generation QLED 8K line at CES 2020 in Las Vegas, offering a glimpse into the future of display technology. ...
Senior Member
Posts: 17563
Joined: 2009-02-25
Later on perhaps unless a earlier deal was already in place, launching these now would be a bit too late for the engineering stage of Navi20 short of a delay though who knows what deals might be in place with AMD and Samsung as one of their primary GPU RAM suppliers so there's a chance perhaps but I'm thinking it's a bit too late for how much time it takes from planning to fabrication and assembly for the hardware and how changes to steps in this will add a longer delay and having to re-test and redo stuff.

Late 2020 or early 2021 perhaps but mid 2020 launches if that's what might be planned for the next Navi GPU's (Nothing is confirmed though far as I know.) eh guess we'll see but there's other HBM2 and variants that would still be a improvement over what Vega had even if it's not these Gen3 chips just yet.

EDIT: Not that I would know but from how I see it and my own opinion on it at least, complications on what seems to be HBM integration and parts like the interposer and bridge to the GPU core wouldn't be a small bit either was even plans of some low cost HBM version I think that didn't use this bit though with some trade-offs as a result but less costly too.
Which I assume is also a major part of HBM or GDDR for just how much of the total GPU cost these memory modules actually take up and thus overall pricing of the card itself to cover for this.
(So more mid-range cards on HBM might still be a bit too early due to the costs and resulting price tag.)
EDIT: Well that and gaming wise as Navi is more oriented towards HBM isn't a immediate benefit over GDDR6 although the GPU design itself could be more or less memory bottlenecked.
Would be interesting to see and have numbers on a HBM Navi GPU though although it wouldn't be a direct comparison if it's high-end/enthusiast cards only like the rumored 5900's if that's reliable.
(But who knows maybe Big Navi could also be compute oriented a bit but it seems AMD has plans for GCN yet.)
Price wise perhaps rather than HBM vs GDDR since HBM has some nice improvements but GDDR is probably not going anywhere anytime soon.
(Eh just speculative for how costly it is and other complications HBM involves even if there are benefits to using it instead of GDDR.)
Senior Member
Posts: 3408
Joined: 2013-03-10
You'd think AMD would stick to HBM, seeing how AMD was one of the developers of HBM in the first place. Whether it's only for professional cards or also gaming, is a different thing.
But for sure these will appear in Nvidia's best of the best professional products. I reckon Nvidia would use HBM in private consumer products only if they simply couldn't make GDDR work with a Titan/x080 Ti (taking also the power consumption into account). Ever since Maxwell, Nvidia has been better at handling GDDR and a narrower memory interface than AMD, though.
Senior Member
Posts: 16438
Joined: 2014-10-25
Do we really need 32GB HBM2?
IMhO 16GB will be plenty enough, even for 4k gaming.
I would prefer 4x stack over 2x stack any time... (4x4GB or 2x8GB or 4x8GB).
Senior Member
Posts: 15745
Joined: 2018-03-21
This will be used on Ampere
Yes
Senior Member
Posts: 22371
Joined: 2008-08-28
Any chance big navi could use this?